1

Suppose I have two std::vectors x and y and a binary function F. I would like to create a new object z (not necessarily a vector), with the property that the ith elemenent of z will be the application of F to the ith elements of x and y. For the time being I use 

boost::range::transform(x, y, z.begin(), F)

which of course requires memory allocation for z. My goal however is to avoid this memory allocation and have the transformation lazy evaluated, similarly to what boost::adaptors::transformed does with unary functions. In other words, the type of z would be some type of boost::range. Is there any easy way to do this or would I have to write my own range class? Thanks in advance!

TemplateRex
  • 69,038
  • 19
  • 164
  • 304
linuxfever
  • 3,763
  • 2
  • 19
  • 43
  • @maverik: thanks for your answer. Unfortunately, that will not achieve my goal since memory would still need to be allocated for z (this time sequentially). – linuxfever Jun 07 '13 at 11:26
  • Even if one could achieve lazy evaluation, how do you imagine avoiding memory allocation when the vector size is unknown at compile time? I can only think of reserving enough memory once instead of sequentially; and depending on the context, one could leverage `alloca()`. Are those two possibilities an option? – klaus triendl Oct 30 '13 at 21:14

1 Answers1

1

The problem is that Boost.Range does not provide a zip_range that can zip x and y as the inputfor your transform. So you need to write one yourself, e.g. using make_iterator_range and zip_iterator (from the Boost.Iterator library). See e.g. this answer for a code example.

Community
  • 1
  • 1
TemplateRex
  • 69,038
  • 19
  • 164
  • 304